Infrared small target detection method and device based on similarity measurement of adjacent blocks
By generating support and contrast regions on infrared images, calculating similarity for estimation, and subtracting the original image from the generated background estimation image, the problem of inaccurate estimation in infrared small target detection under complex backgrounds is solved, achieving higher detection accuracy and lower false alarm rate.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- WUHAN HUAZHONG TIANYI INTELLIGENT TECH CO LTD
- Filing Date
- 2023-10-27
- Publication Date
- 2026-07-21
AI Technical Summary
In existing technologies, infrared small target detection is inaccurate in complex backgrounds, resulting in a high false alarm rate.
By generating a geometric support region around the pixel to be estimated on the infrared image, and generating several contrast regions around it, the pixel in the contrast region with the same position as the pixel to be estimated is calculated and estimated. After generating a background estimation image, the original infrared image is subtracted to separate the infrared small target.
It effectively eliminates the influence of complex backgrounds on the detection of small infrared targets, improves the accuracy of estimation, and reduces the false alarm rate.
Smart Images
Figure CN117557773B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of image processing technology, and in particular to an infrared small target detection method and apparatus based on neighbor block similarity measurement. Background Technology
[0002] Infrared imaging can convert infrared signals of different bands emitted by an object into images that can be distinguished by the human eye, and is therefore widely used in various fields such as long-distance remote sensing and video surveillance. Due to the long imaging distance, the observed target usually appears as a patch or dot in the infrared image, with an image size of only a few to tens of pixels. This results in a low signal-to-noise ratio and a lack of effective features such as shape and color. Traditional filtering-based infrared background estimation algorithms establish a support domain of a specific size centered on the pixel to be estimated. Then, using the gray values of the pixels within the support domain, different filtering algorithms are applied to estimate the pixel, such as median filtering based on the pixel median, Gaussian weighted filtering, and mean filtering based on the pixel mean. These algorithms mainly use the gray values of pixels within the support domain to estimate the background. When the pixel to be estimated is located at a complex edge or corner, the difference between the pixels at the edge and corner and the local gray values of the support domain can be significant. Therefore, considering only the correlation between the pixel to be estimated and other pixels within the support domain in terms of gray values is inaccurate for pixels in complex locations, leading to a high false alarm rate. Summary of the Invention
[0003] This application provides an infrared small target detection method based on neighbor block similarity measurement, which solves the technical problem of inaccurate estimation of pixels in complex locations in the prior art.
[0004] To achieve the above objectives, the technical solutions adopted in the embodiments of this application are as follows:
[0005] In a first aspect embodiment, this application provides an infrared small target detection method based on neighbor block similarity measurement, the method comprising:
[0006] Receive infrared images;
[0007] A geometric support region is generated around the pixel to be estimated on the infrared image;
[0008] Several contrast areas with the same shape and size as the support area are generated around the support area;
[0009] Calculate the similarity between several of the comparison regions and the support region;
[0010] The pixel to be estimated is estimated by using the pixel in the comparison region with the highest similarity to the pixel to be estimated, and a grayscale estimate is obtained;
[0011] The above-described method for estimating the pixels to be estimated is used to estimate all pixels in the infrared image to obtain grayscale estimates of all pixels and generate a background estimation image.
[0012] The target image is obtained by subtracting the infrared image from the estimated background image.
[0013] The target is separated from the target image to complete the infrared small target detection.
[0014] This application generates several contrast regions around a support region, calculates the similarity between the contrast regions and the support region, and uses the pixel in the contrast region with the highest similarity that has the same coordinate position as the pixel to be estimated to estimate the pixel, thus obtaining a grayscale estimate. After estimating the grayscale values of all pixels in the infrared image, a background estimation image is generated. The target image is obtained by subtracting the background estimation image from the original infrared image. Then, the infrared small target region in the target image is separated to complete the infrared small target detection. This application uses the pixel in the contrast region with the highest similarity to the support region that has the same coordinate position as the pixel to be estimated during the estimation process, which effectively eliminates the influence of complex backgrounds on infrared small target detection and solves the technical problem of inaccurate estimation of pixels in complex positions in the prior art.
[0015] In some embodiments, the support region is rectangular in shape.
[0016] In some implementations, the process of generating the comparison region includes:
[0017] A search area is generated centered on the support area;
[0018] The search area, excluding the support area, is divided into several sub-regions with the same shape and size as the support area. These sub-regions are the comparison areas.
[0019] In some implementations, the similarity calculation process includes:
[0020] Calculate the maximum grayscale value of the support area;
[0021] Calculate the average gray value of the contrast region;
[0022] The local contrast is obtained by multiplying the maximum gray value of the support area by the ratio of the maximum gray value of the support area to the average gray value of the contrast area;
[0023] The reciprocal of the local contrast is calculated to obtain the similarity.
[0024] In some implementations, the expression for the local contrast is:
[0025]
[0026] Among them, L n m represents the maximum grayscale value of the supporting region. i This represents the average gray value of the i-th contrast region surrounding the support region.
[0027] In some implementations, the similarity calculation process includes:
[0028] The pixels in the support region are designated as first pixels, and the pixels in the contrast region are designated as second pixels. The sum of the Euclidean distances between all the first pixels and the second pixels corresponding to the positions of all the first pixels is calculated to obtain the local contrast.
[0029] The reciprocal of the local contrast is calculated to obtain the similarity.
[0030] In some implementations, the expression for the local contrast is:
[0031]
[0032] Where (x0,y0) and (x k ,y k ) is the supporting region v0 and the contrast region v k The coordinates of the pixels with the same position as the pixel to be estimated, I(x0,y0) and I(x k ,y k ) is its corresponding grayscale value, and the dimensions of the support area and the contrast area are (2v+1, 2v+1).
[0033] In some implementations, an adaptive thresholding segmentation algorithm is used to separate the target from the target image, as shown in the following expression:
[0034] Th=u+k*σ,
[0035] Where Th represents the threshold, u represents the mean gray level, σ represents the standard deviation of gray level, and k is the variance coefficient;
[0036] A pixel in the target image is marked as a target when its grayscale value is greater than the threshold value.
[0037] In a second aspect of the embodiment, this application provides an infrared small target detection device based on neighbor block similarity measurement, comprising:
[0038] The receiving module is used to receive infrared images;
[0039] A support region generation module is used to generate a geometric support region around the pixel to be estimated on the infrared image.
[0040] A contrast region generation module is used to generate several contrast regions around the support region that have the same shape and size as the support region.
[0041] The first data processing module is used to calculate the similarity between the supporting region and several of the comparison regions respectively;
[0042] The second data processing module is used to estimate the pixel to be estimated by using the pixel with the same coordinate position as the pixel to be estimated in the comparison region with the highest similarity, so as to obtain the grayscale estimate value.
[0043] The third data processing module is used to estimate all pixels in the infrared image, obtain the grayscale estimate of all pixels, and generate a background estimation image;
[0044] The fourth data processing module is used to subtract the background estimation image from the infrared image to obtain the target image;
[0045] The separation module is used to separate the target from the target image to complete the infrared small target detection.
[0046] Additional aspects and advantages of embodiments of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of embodiments of this application. Attached Figure Description
[0047] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0048] The methods, systems, and / or procedures shown in the accompanying drawings will be further described with reference to exemplary embodiments. These exemplary embodiments will be described in detail with reference to the drawings. These exemplary embodiments are non-limiting exemplary embodiments, wherein example figures represent similar mechanisms in the various views of the drawings.
[0049] Figure 1 This is a flowchart of an infrared small target detection method based on neighbor block similarity measurement in one embodiment;
[0050] Figure 2 This is a flowchart of the process of generating the comparison region in one embodiment;
[0051] Figure 3 This is a schematic diagram of the support area, comparison area, and search area in one embodiment;
[0052] Figure 4 This is a flowchart of the similarity calculation process in one embodiment;
[0053] Figure 5 This is a flowchart of another similarity calculation process in one embodiment;
[0054] Figure 6 This is a schematic diagram illustrating the process of estimating the pixel to be estimated in one embodiment;
[0055] Figure 7 This is a schematic diagram of the process of obtaining the target image in one embodiment;
[0056] Figure 8 This is a schematic diagram illustrating the visualization effect after performing full-image similarity calculations on the regions where infrared small target pixels are located and the regions where edge pixels are located, respectively, in one embodiment;
[0057] Figure 9 This is a structural block diagram of an infrared small target detection device based on neighbor block similarity measurement in one embodiment. Detailed Implementation
[0058] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.
[0059] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.
[0060] In the description of this invention, it should be noted that the terms "first," "second," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.
[0061] Please see Figure 1 In a first aspect embodiment, this application provides an infrared small target detection method based on neighbor block similarity measurement, comprising:
[0062] S1. Receive infrared images.
[0063] S2. Generate a geometric support region around the pixel to be estimated on the infrared image.
[0064] In some implementations, the support area is rectangular in shape.
[0065] It should be noted that the shape of the support area can be any geometric shape, such as triangle, hexagon, parallelogram, circle, etc. Since the infrared images acquired by the acquisition device are mostly rectangular, this application uses a rectangle as the shape of the support area to facilitate the processing of the edge parts of the infrared image.
[0066] S3. Generate several contrast areas around the support area that have the same shape and size as the support area.
[0067] Please see Figure 2 In some embodiments, the process of generating the comparison region includes:
[0068] S31. Generate a search area centered on the support area.
[0069] Please see Figure 3 The area within the red box is the support region, the area within the blue box is the comparison region, and the area within the yellow box is the search region. It should be noted that the size of the search region can be adjusted according to actual application requirements. To improve the accuracy of this method, the entire infrared image could be used as the search region, but this would significantly increase the algorithm's time complexity and severely impact its performance. The search region generated in this step determines the range of the comparison region. Subsequently, only the comparison region with the highest similarity to the support region will be searched within the search region. Analysis of the characteristics of infrared images reveals that the support region of any pixel, as a local region of any structure in the infrared image, always has a high similarity to its surrounding neighboring regions. However, as the distance increases, the structural differences between the support region and more distant regions become more pronounced, and the similarity decreases. Therefore, in actual calculations, it is only necessary to take a certain size range around the support region as the search region. Therefore, to avoid excessive time complexity and severely impacting algorithm performance, this embodiment selects the size of the search region to be 3 or 4 times the size of the support region.
[0070] Preferably, to avoid the influence of the similarity of the support region itself and reduce the bias caused by infrared small targets in estimating the background image, the contrast region adjacent to the support region is excluded during similarity calculation, i.e. Figure 3 The area between the red and blue boxes.
[0071] S32. Divide the search area excluding the support area into several sub-regions with the same shape and size as the support area. These sub-regions are the comparison areas.
[0072] S4. Calculate the similarity between the comparison region and the support region respectively.
[0073] Please see Figure 4 In some embodiments, the similarity calculation process includes:
[0074] S411. Calculate the maximum gray value of the support area.
[0075] S412. Calculate the average gray value of the contrast region.
[0076] S413. Multiply the maximum gray value of the supporting region by the ratio of the maximum gray value of the supporting region to the average gray value of the contrast region to obtain the local contrast.
[0077] S414. Calculate the reciprocal of the local contrast to obtain the similarity.
[0078] In some implementations, the expression for local contrast is as follows:
[0079]
[0080] Among them, L n This represents the maximum grayscale value of the supported region, m. i This represents the average gray value of the i-th contrast region surrounding the support region.
[0081] It should be noted that the calculation methods for the maximum and average gray values within a certain area of an image are existing technologies, and therefore will not be elaborated here.
[0082] Please see Figure 5 It should be noted that the local contrast calculated using methods S411 to S413 amplifies the contrast between high-brightness noise and imager defects in the infrared image and their surrounding areas, potentially causing the target pixel to be submerged in noise. To address this issue, in the following embodiments, this application uses an adaptive local contrast measurement algorithm to calculate the local contrast and thus obtain the similarity, including:
[0083] S421. Label the pixels in the support region as the first pixel and the pixels in the contrast region as the second pixel. Calculate the sum of the Euclidean distances between all the first pixels and the second pixels corresponding to the positions of all the first pixels to obtain the local contrast.
[0084] S422. Calculate the reciprocal of the local contrast to obtain the similarity.
[0085] The adaptive local contrast measurement algorithm no longer only calculates the average and highest gray values of pixel blocks, but also fully considers the structural characteristics of the target, calculating the contrast down to each individual pixel and summing the Euclidean distances between corresponding pixels in different regions. Therefore, even if a pixel has high contrast due to noise or bad pixels, it has no significant impact on the contrast of the entire region.
[0086] In some implementations, the expression for local contrast in the adaptive local contrast measurement algorithm is:
[0087]
[0088] Where (x0,y0) and (x k ,y k ) is the supporting region v0 and the kth comparison region v k The coordinates of the pixels with the same position as the pixel to be estimated, I(x0,y0) and I(x k ,y k ) is its corresponding grayscale value, and the size of the support area and the contrast area is (2v+1, 2v+1).
[0089] The greater the difference in adaptive local contrast between regions, the greater the difference in grayscale values of their internal pixels, and the smaller the similarity. Therefore, the similarity between the supporting region v0 and its surrounding k-th contrast region can be obtained by the following formula:
[0090]
[0091] Clearly, the similarity s between regions k With support region v0 and contrast region v k The adaptive local contrast is inversely proportional. When v0 and v k When the grayscale values are exactly the same, the adaptive local contrast between the two regions is zero. k The maximum value is 1. v0 and v k The greater the difference in grayscale values, the more s k The smaller the value, the less similar the regions are.
[0092] S5. Using the pixel in the comparison region with the highest similarity that has the same coordinate position as the pixel to be estimated, estimate the pixel to be estimated and obtain the grayscale estimate.
[0093] The comparison region V with the highest similarity to the comparison region V0 s It can be expressed by the following formula:
[0094]
[0095] V s =Vi ,
[0096] Among them, S i This represents the similarity between the supporting region and the i-th surrounding comparison region.
[0097] In some embodiments, based on the most similar comparison region V s The grayscale value of the pixel with the same coordinate position as the pixel to be estimated is used to obtain the grayscale estimate of the pixel (x0, y0) using a filtering algorithm. Considering that the pixel to be estimated (x0, y0) and the contrast region V with the highest similarity s The pixel whose coordinates are the same as the pixel to be estimated (x) s ,y s Pixels belonging to the same structure are more likely to have the same grayscale value. Therefore, this application uses inverse gradient weighting to estimate the grayscale value of the pixel to be estimated (x0, y0), and the expression for the grayscale estimate is as follows:
[0098]
[0099] in, V represents the grayscale estimate, (x0, y0) is the coordinate of the pixel to be estimated, and V s The region with the highest similarity is (x s ,y s ) is V s The coordinates of the pixel whose coordinates are the same as the pixel whose coordinates are to be estimated, I(x) s ,y s ) represents V s The grayscale value of the pixel whose coordinates are the same as the pixel to be estimated is given by the following formula: The parameter W is the weight normalization parameter, which is defined as follows:
[0100]
[0101] S6. Estimate the grayscale values of all pixels in the infrared image to generate a background estimation image.
[0102] Please see Figure 6In the image, the first column from the left is the infrared image, where (x, y) represents the pixel to be estimated. The red box in the second column represents the support region, and the blue box represents the contrast region. The fourth column shows the background estimation image of the support region obtained after estimating the pixel (x, y) and replacing its original grayscale value with its estimated grayscale value. It can be seen that the background estimation image of the support region has filtered out the bright pixel (x, y) to be estimated. It should be noted that the grayscale values of each pixel in the background estimation image generated in step S6 are the grayscale estimates of the pixels in the original infrared image corresponding to those pixel positions.
[0103] S7. Subtract the estimated background image from the infrared image to obtain the target image.
[0104] It should be noted that subtracting the infrared image from the background estimation image means subtracting the grayscale value of each pixel in the background estimation image corresponding to the given coordinate position from the grayscale value of each pixel in the infrared image. Please refer to [link to relevant documentation]. Figure 7 , Figure 7 (a) is the original infrared image. Figure 7 (b) is the background estimation image. Figure 7 (c) is the target image. Figure 7 (c) shows the infrared small target within the red box. It can be seen that the brightness of the infrared small target in the target image is significantly higher than that of other areas, which facilitates subsequent separation operations.
[0105] S8. Separate the target from the target image to complete the infrared small target detection.
[0106] In some embodiments, an adaptive threshold segmentation algorithm is used to separate the target from the target image. The adaptive threshold segmentation calculation formula is as follows:
[0107] Th=u+k*σ
[0108] The threshold Th is determined by the mean gray level u and the standard deviation σ of the target image, and k is the variance coefficient. Pixels in the target image with a gray level greater than Th are marked as targets.
[0109] Please see Figure 8The first column from the left in the image shows two infrared images. The red box contains small infrared targets, and the blue box contains edge regions. Since both edge regions and small infrared targets appear as areas composed of bright pixels in infrared images, traditional pixel estimation methods only consider the difference in grayscale values between the pixel to be estimated and other pixels within the support region. This can easily lead to edge regions being marked as targets, resulting in inaccuracies. The infrared small target detection method based on neighbor block similarity measurement provided in this application generates several contrast regions around the support region and finds the one with the highest similarity to the support region. The pixel to be estimated is then estimated using this contrast region. Figure 8 The second column is a visualization of the similarity calculation between the area within the red box and the entire image, and the third column is a visualization of the similarity calculation between the area within the blue box and the entire image. The higher the grayscale value, the higher the similarity. It can be seen that since the infrared small target only has high similarity to itself, the red box cannot find any other similar areas besides itself. Therefore, when estimating pixels later, the estimated grayscale value will differ significantly from the pixel's actual grayscale value. The infrared small target area formed in the target image will also have a significant grayscale difference from other areas, allowing for accurate labeling. The edge region not only has a high similarity to itself, but also has a high similarity to regions belonging to the same edge. Therefore, when estimating pixels in the subsequent process, the estimated gray value of pixels at the edge will not differ much from the actual gray value of the pixel. In the subsequent generated background estimation image, it will also exist as a highlight and will be subtracted as background without affecting the target image. Pixels at corner positions have similar properties to pixels at the edge positions, and this method can also accurately identify them. Therefore, the infrared small target detection method based on neighbor block similarity measurement provided in this application can accurately estimate pixels in complex positions, solve the problem of inaccurate estimation of pixels in complex positions in the prior art, and reduce the false alarm rate.
[0110] Please see Figure 9In a second aspect embodiment, this application provides an infrared small target detection device based on neighbor block similarity measurement, including a receiving module, a support region generation module, a contrast region generation module, a first data processing module, a second data processing module, a third data processing module, a fourth data processing module, and a separation module. The receiving module receives an infrared image. The support region generation module generates a geometric support region around the pixel to be estimated in the infrared image. The contrast region generation module generates several contrast regions with the same shape and size as the support region around it. The first data processing module calculates the similarity between the support region and the several contrast regions. The second data processing module estimates the pixel to be estimated using the pixel with the same coordinate position as the pixel to be estimated in the contrast region with the highest similarity, obtaining a grayscale estimate. The third data processing module estimates all pixels in the infrared image to obtain a background estimation image. The fourth data processing module subtracts the background estimation image from the infrared image to obtain a target image. The separation module separates the target from the target image, completing the infrared small target detection.
[0111] It is understandable that the various modules in the infrared small target detection device based on neighbor block similarity measurement can be integrated on a processor, or each module can exist separately in different processors, or two or more modules can be integrated into a processor. In addition to being implemented in hardware such as processors, these modules can also be implemented in the form of software functional modules.
[0112] Each module of the infrared small target detection device based on neighbor block similarity measurement can execute the corresponding steps in any of the infrared small target detection methods based on neighbor block similarity measurement mentioned above. Therefore, the details of each module will not be elaborated here. Please refer to the corresponding step descriptions above for details.
[0113] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application, the scope of which is defined by the claims and their equivalents.
Claims
1. An infrared small target detection method based on neighbor block similarity measurement, characterized in that, include: Receive infrared images; A geometric support region is generated around the pixel to be estimated on the infrared image; Several contrast areas with the same shape and size as the support area are generated around the support area; The process of generating the comparison region includes: A search area is generated centered on the support area; The search area excluding the support area is divided into several sub-regions with the same shape and size as the support area, and the sub-regions are the comparison areas; Calculate the similarity between several of the comparison regions and the support region; exclude comparison regions adjacent to the support region during the similarity calculation. The pixel to be estimated is estimated by using the pixel in the comparison region with the highest similarity to the pixel to be estimated, and a grayscale estimate is obtained; The above-described method for estimating the pixels to be estimated is used to estimate all pixels in the infrared image to obtain grayscale estimates of all pixels and generate a background estimation image. The target image is obtained by subtracting the infrared image from the estimated background image. The target is separated from the target image to complete the infrared small target detection.
2. The infrared small target detection method based on neighbor block similarity measurement as described in claim 1, characterized in that, The supporting area is rectangular in shape.
3. The infrared small target detection method based on neighbor block similarity measurement as described in claim 1, characterized in that, The similarity calculation process includes: Calculate the maximum grayscale value of the support area; Calculate the average gray value of the contrast region; The local contrast is obtained by multiplying the maximum gray value of the support area by the ratio of the maximum gray value of the support area to the average gray value of the contrast area; The reciprocal of the local contrast is calculated to obtain the similarity.
4. The infrared small target detection method based on neighbor block similarity measurement as described in claim 3, characterized in that, The expression for the local contrast is: ， in, This represents the maximum grayscale value of the supporting region. This represents the average gray value of the i-th contrast region surrounding the support region.
5. The infrared small target detection method based on neighbor block similarity measurement as described in claim 1, characterized in that, The similarity calculation process includes: The pixels in the support region are designated as first pixels, and the pixels in the contrast region are designated as second pixels. The sum of the Euclidean distances between all the first pixels and the second pixels corresponding to the positions of all the first pixels is calculated to obtain the local contrast. The reciprocal of the local contrast is calculated to obtain the similarity.
6. The infrared small target detection method based on neighbor block similarity measurement as described in claim 5, characterized in that, The expression for the local contrast is: ， in, and The supporting area and the comparison area The coordinates of the pixels in the same position as the pixel whose coordinates are to be estimated. and It is its corresponding grayscale value, and the dimensions of the support area and the contrast area are... .
7. The infrared small target detection method based on neighbor block similarity measurement as described in claim 1, characterized in that, The expression for the grayscale estimate is as follows: ， in, This represents the estimated grayscale value. The coordinates of the pixel to be estimated are... The region with the highest similarity is the comparison region. for The coordinates of the pixels in the same position as the pixel whose coordinates are to be estimated. express The grayscale value of the pixel whose coordinates are the same as the pixel to be estimated, and the parameters. The weight normalization parameter is defined as follows: 。 8. The infrared small target detection method based on neighbor block similarity measurement as described in claim 1, characterized in that, An adaptive thresholding segmentation algorithm is used to separate the target from the target image, as shown in the following expression: ， in, Indicates the threshold. Indicates the grayscale mean. Indicates the standard deviation of gray levels. The variance coefficient; A pixel in the target image is marked as a target when its grayscale value is greater than the threshold value.
9. The apparatus for infrared small target detection based on neighbor block similarity measurement as described in any one of claims 1-8, characterized in that, include: The receiving module is used to receive infrared images; A support region generation module is used to generate a geometric support region around the pixel to be estimated on the infrared image. A contrast region generation module is used to generate several contrast regions around the support region that have the same shape and size as the support region. The first data processing module is used to calculate the similarity between the supporting region and several of the comparison regions respectively; The second data processing module is used to estimate the pixel to be estimated by using the pixel with the same coordinate position as the pixel to be estimated in the comparison region with the highest similarity, so as to obtain the grayscale estimate value. The third data processing module is used to estimate all pixels in the infrared image, obtain the grayscale estimate of all pixels, and generate a background estimation image; The fourth data processing module is used to subtract the background estimation image from the infrared image to obtain the target image; The separation module is used to separate the target from the target image to complete the infrared small target detection.